There was a recent funny article on “How to be the Perfect Mother” from Huffington Post that was a hilarious look at how society tells us conflicting information about how we should act as mothers. You should go look at it if you are a mother, know a mother, or have a mother. Just go see it.
This article, combined with two recent manuscript reviews coming back, got me thinking about how reviewers also often write conflicting advice for your manuscripts. So, I decided to write a satirical version of a manuscript review as an example.
***Note: any resemblance to reviews you may have received or written are purely coincidental.
We have read and reviewed the manuscript, “This Science Thing is Important for this Other Thing” by Prof.Science. This manuscript investigates the ScienceThing and its interactions with OtherThing, a very important and understudied topic. This group performed many new experiments that had never been done before and had 6 figures each with A-J panels. Their work was executed well and revealed new information about the interactions of ScienceThing with OtherThing that we never knew before. Their clearly written manuscript had a simulation that modeled the results and showed similar trends suggesting a mechanism.
In performing these experiments, they used well-tested experimental methods along with specific tests to control for errors. They have used these methods to test for effects of ScienceThing on OtherThing and have quantified the effects. Since these methods are well-tested and accepted in the field, they are not novel. We want only novel experiments even if we cannot interpret the results we get from them. Thus, we suggest that the authors perform all new experiments. Further, did the authors investigate how ScienceThing affected OtherThingII? Only one paper on OtherThingII exists, from the OldFart Group, but it is clearly more important than OtherThing, and it should be explored even though almost no reagents exist for OtherThingII. Unless OtherThingII is also investigated, I do not think this paper is very worthwhile.
The authors display histograms of their work and how ScienceThing affects the OtherThing. It is important to be quantitative and have numerical data. For each histogram, they fit to a Gaussian and report the R-squared value of the fit to the data. They use these fits to discuss the results. Why do they do this? Why not use a simple p-value to the data? Isn’t a student’s t-test done on everything? It is clear that the two distributions do not overlap, so they should report the p-value.
The authors used a toy model to show that the ScienceThing behavior that they see could be due to a minimal number of simple rules. Being quantitative and having models is important. We want more quantitative work and models in this field of science. The simulation has the same trends as the experimental data, but it does not exactly match the data, so the model must be worthless. Why did these authors have a model? They are not theorists or modelers; they are experimentalists. They should remove the model, it detracts from the data.
Without the model, the authors do not have a mechanism. We want all science to be mechanistic. It is not good enough to simply observe something and report what happens. For instance, although their toy model uses 3 simple rules and has the same general trends as the data, they cannot rule out a model with 10 complicated rules. Thus, they have not revealed the mechanism behind the results they see, and thus the impact of the work is lower in my opinion. Until their work becomes more mechanistic, their results are purely qualitative, and the work is not work publishing.
There are a number of sp errs in this manuscript. Don’t they care how they present thmseves? Its like thei didn’t even porrof read before they sent it out. They need to really fix this. There are way too many issues for me to helpfully point out.
They are missing a number of very important citations particularly from the OldFart group, “Science Stuff: A novel Regulator of Nothing,” JSS 1979; “Science Stuff Moves Science Thing,” JSS 1998; and “Science Stuff to Science Thing,” AJSS 2000. These important references about how ScienceStuff is connected to ScienceThing are important and should be added.
Their experimental methods are not good. They didn’t even present them! I suppose they could be in the supplement, but I didn’t read it, so I wouldn’t know. Even if they are in the supplement, they need to have them in the main text. Maybe, once they take out the model, they will have room in this 5-page paper to have detailed methods.
In conclusion, after having read this paper, I feel that these results were obvious and could have been guessed from deductive reasoning. Thus, the experiments were not necessary and the results are not novel. Further, to make the results important and novel, the authors would need to perform a number of extra experiments that were not in the original 60 plots presented, and they would need a mechanism, which they have not proven. Overall, it is clear that this study has no value and, thus, I recommend that this paper be rejected.
Anything to add? Post or comment here. Maybe we can add more examples? If you want to get an email every time I post, push the +Follow button.